sid pfa status and calorimeter performance ron cassell (slac) sid design study meeting 11/15/08

SiD PFA Status and Calorimeter Performance

Ron Cassell (SLAC)

SiD Design Study Meeting 11/15/08

2

Overview

• PFA goals for the LOI

• Progress since Boulder

• Current performance

• Use in benchmarking analysis

• The LOI

3

From Mat at Boulder

4

PFA goals for the LOI

• A stable reconstruction program: -> Output reconstructed particles to be used for analyses of LOI benchmark processes.

• To be run on full SM and data sample.

• Improvements, bug fixes, etc. may warrant rerunning full sample, executive decision will be needed.

5

Again from Mat

6

Progress

• UI PFA completely refactored: Code reorganized to be maintainable, critical with Mat’s departure.

• Muon hits handled in a consistent way (although probably not optimal)

• First pass lepton ID• Full tracking now the default• Production release of the lcsim package• Output usable by benchmarking group• Fixed error in running FastMC on simulated data• Critical decisions: sid02 is the default detector,

and full tracking will be used.

7

Current performance

• Benchmarking analyses are what count!• PFA tests are what is shown.• In following slides, Prod == sid02, full tracking. (no

cheating)• For comparisons, PPR == perfect pattern recognition

(cheat on tracking, cheat on calorimeter hit assignments)• FastMC == Fast Monte Carlo (Use pythia final state

particles with smearing, tuned to give Pandora-like results for a super-detector.

• CalOnly == pure calorimeter energy measurement.

8

qq(uds) events at fixed Ecm

rms90 = 3.63 GeV

sigma = 4.67 GeV

rms = 5.14 GeV

rms90 = 9.44 GeV

sigma = 11.8 GeV

rms = 13.9 GeV

rms90 = 6.00 GeV

sigma = 7.65 GeV

rms = 8.27 GeV

rms90 = 16.6 GeV

sigma = 20.1 GeV

rms = 26.2 GeV

9

qq(uds) events at fixed Ecm

10

ZZ events at 500 GeV, max cos(theta) < 0.95

• Full rms = 5.71 GeV

• Sigma(gauss) = 5.11 GeV

• rms90 = 4.00 GeV

• (dM/M)90 = 4.48%

11

ZZ events at 500 GeV, max cos(theta) < 0.95

• Sigma(gauss) = 6.19 GeV

12

Current performance• I could stop here. The previous slides are the current status of PFA

development.• Try to put in perspective by comparisons.• CalOnly – using only the calorimeters (no tracking) what is the

energy resolution for sid02?• Cheat tracking – quantify resolution loss using full tracking package.• PPR – the potential of Pflow: if we could only make perfect

associations.• FastMC – our only real connection to physics output vs detector

design. Since most of the analyses are/were being developed with FastMC, comparison of results with PFA package may help quantify energy resolution -> physics results.

• What about scintillator? And Pandora?

13

Comparison of CalOnly and Prod Event energy resolution

14

Comparison of cheat vs real tracking

• Energy resolution worse by 6-7% for Ecm < 200 GeV

• Mass resolution worse by 9%, mainly due to barrel region

• Full tracking has pt cut (>200MeV) and impact parameter cut. No kink reconstruction or tracks from vees.

• Marcel once reported (from Mark Thompson) that kink and vee reconstruction improved resolution ~ 5%.

• Excellent result!

(dM/M)90

Cheat tracking Full tracking

Barrel 4.28% 4.73%

Forward 3.72% 3.96%

Both 4.04% 4.33%

Combined 4.08% 4.45%

15

Comparison of PPR and Prod reconstruction

• We see why the emphasis on pattern recognition

PPR Prod

RMS90 2.24 GeV 4.00 GeV

dM/M 2.46% 4.48%

16

Comparison of FastMC and Prod reconstruction

• Interesting that FastMC gives better mass resolution than PPR, with much worse energy resolution.

FastMC Prod (PPR)

RMS90 2.01 GeV 4.00 GeV (2.24 GeV)

dM/M 2.23% 4.48% (2.46%)

17

Jet energy resolution

• dE/E = alpha/sqrt(E)

18

Current performance caveats

• Mat reported at Boulder similar performance for low energy jets as pandora … using sid01_scint, cheat tracking, and comparing to Marcel’s sidish detectors. But …

• Scint -> rpc ~ 10% worse jet energy resolution. Cheat tracking to real tracking -> 7% worse jet energy resolution.

• This is where we are.

19

Lepton ID

• The test samples have no prompt leptons. (uds quarks, no neutrinos)

• Need physics processes to test lepton ID.

20

Muons

• Code from Tae Jeong did not make it into production release. (Many places it should have been caught, can spread blame around).

• Prod reconstruction output has no identified muons.

• Problems have been identified and fixed, and a post-reconstruction identifier is being developed.

21

22

Electrons

• Tim looked at electrons in the ZH sample, and found only ~50% efficiency for electrons > 20 GeV.

• He has developed an algorithm for identifying these electrons, again can be done post-reconstruction.

23

Benchmark analyses

• Reconstruction output seems “suitable”.• Feedback from benchmarking group has been essential

to get to this point, and is even more critical now as LOI deadline nears.

• Comparison of FastMC with Prod may well guide us in

post LOI detector optimization.“suitable”: In ttbar events, the reconstruction output has been run through vertexing/flavor tagging process, yielding a mass plot.

24

25

Summary

• We have a full reconstruction package with NO cheating.• SM and data sample processing well under way.• Original performance goals, even with caveats, not quite

met (We’re not out of ideas, we’re out of time). Actual performance on test samples has been documented.

• Benchmark analyses starting to use PFA output.• Since a second reconstruction pass is far from given,

should now concentrate on improvements that can be applied post-reconstruction.